Professional Documents
Culture Documents
5 (2009)
PAGES **-**
ABSTRACT
Existing reinforced concrete buildings in north India were built at times when seismic zones
were not recognized. These buildings need to monitor the damage and improve their load
carrying capacities for the purpose of seismic safety. This paper investigates the effect of
lateral load on the damage index of reinforced concrete frame model strengthened using
FRP sheets. Stiffness damage index and combined damage parameters approach were used.
Change in modal parameters methods were investigated with the help of impact hammer
excitation test. Finally, finite element method was used for modelling, nonlinear analysis
and results processing of the specimen.
Keywords: Damage index, retrofitted RC frame model, FRP sheet, nonlinear FE method
1. INTRODUCTION
In north India, existing reinforced concrete (RC) buildings were constructed at time when
seismic zones were not recognized. Therefore, they have inadequate reinforcement detailing
which not results in deficient lateral load resistance. For the purpose of assuring seismic
safety, it is necessary to detect the damaged state. Kanwar et.al [1] found that non-ductile
RC moment resisting frames result in the need of retrofitting to increase the load resistance
capacities. There are various methods of strengthening the damaged structures. The use of
steel plate jacket and ferro-cement jacket are disruptive to the operation of the facility, labor
intensive and time consuming. The Fiber Reinforcement Polymer (FRP) is one popular
strengthening technique because FRP with epoxy have received considerable attention due
to its high strength, light weight, quick and easy manageability on-site and high resistance
against corrosion. This advance material has been successfully used to increase bending and
shear capacity of flexural elements [2,3]. The retrofitted columns using FRP sheets have
shown to prevent its brittle shear failure and also significantly improved their displacement
ductility and energy dissipation capacity [4]. The use of FRP systems have also found to be
an effective method for upgrading deficient RC connections [5].
*
Email address of the corresponding author: atiwat.vimut@gmail.com (A. Vimuttasoongviriya)
564 A. Vimuttasoongviriya, N. Kwatra and M. Kumar
For the purpose of structural damage assessment, it is necessary to monitor the structures
for location and extent of damage. Damage index is a quantity that is used for estimating the
damage status. This index depends on different specific damage parameters such as
deformation, structural stiffness, energy dissipation and dynamic properties of structure. The
well known combined damage index method is proposed by Park and Ang [6]. This index is
calculated as a linear combination of maximum displacement response and total hysteretic
energy dissipation under cyclic load. Powell and Allahabadi [7] introduced a deformation
based damage index in terms of displacement or ductility. Kanvar et al. [8] suggested the
method of stiffness damage assessment wherein the damage indicator is based on changes in
the stiffness. Damage detection by calculating the change in modal parameters has been
widely applied in damage alarming in health monitoring systems of highway bridges [9].
Dipasquale and Cakmak [10] presented damage index based on the change ratio of
frequency. This method considers structural fundamental natural frequencies before and
after damaged. Rodriguez and Barroso [11] and Wang et al. [12] proposed stiffness-mass
ratios damage theory. This damage index is based on stiffness-mass ratio in form of modal
parameters. The Modal flexibility damage index method is the well known one [13]. The
principle of this method is on the basis of the comparison of the flexibility matrices obtained
from two sets of mode shapes.
Response of whole building subjected to external load is a problem which is important to
understand and there is little point in performing analysis without testing. This paper intends
to investigate both damage indexes and effects on retrofitted RC frame model under quasi-
static loads. Effect of FRP wraps on RC elements has been reported. The first part involves
determination of load-deformation relationships. Change in stiffness, displacement ductility
and energy dissipation have been investigated. Some of the damage indexes have been
selected to indicate the damage for both nonlinear static and dynamic damage parameters
approach. Subsequently, the ATENA-3D finite element program has been used for
modelling, nonlinear analysis and results processing of the specimen. Three-dimensional
finite element (FE) model is developed to replicate the experimental RC frame before and
after FRP strengthening. Modeling methodology and nonlinear analysis approach in
ATENA-3D are presented.
In structural damage detection, a quantity of damage index (DI) is used to monitor the
damage status of structure. This value is equal to zero when there is no damage and is equal
to one when total collapse occurs. There are various parameters of physical responses to
formulate damage index of structure. These damage parameters can be classified as
deformation, change in stiffness, energy dissipation and changes in dynamical parameters
[14]. A damage index can involve a combination of one or more damage variables in its
calculation. For classification from difference viewpoints, the local damage index indicates
the damage of structural elements or a part of structure and the global damage index
considers the damage of the whole structure. Some of more significant damage indexes were
studied and presented in this paper.
EFFECT OF LATERAL QUASI-STATIC LOAD ON NONLINEAR... 565
2.1 Damage index based on deformation, energy dissipation and change in stiffness
2.1.1 Park and Ang damage index
Park and Ang [6] have formulated a well known damage index to estimate the level of
damage in structures. This damage index contains the damage cause by the maximum
deformation and the damage due to dissipated energy subjected to cyclic loading. Both
components of damage are linearly combined:
δ max β
δ u Py ∫
DI PA = + d Eh (1)
δu
Table 1. Relation between damage index and various damage states [6].
k j − k *j 1
DI k = = 1− (2)
kj νj
566 A. Vimuttasoongviriya, N. Kwatra and M. Kumar
where ν j is stiffness ratio, k j and k *j are the initial stiffness and damage stiffness of the jth
member. The asterisk (*) denotes the damage state.
where ω and ω * are the fundamental eigen frequency and damage frequency parameter,
respectively. In this paper, natural frequencies of structure had been measured from
experimental test in form of frequency response function (FRF) plot. This plot gave FRF
magnitude corresponding to resonance frequency of each mode.
where M, C and K are the system mass, damping and stiffness matrices. x(t) is vector of the
floor displacement relative to the vector of excitation force P(t). From the matrix eigenvalue
problem, whose solution gives the natural frequencies and mode shape of a system, we have:
[K − λ M ]Φ = 0 (5)
where λ is the eigenvalue, and Φ is the mode shape corresponding to λ . The associated
EFFECT OF LATERAL QUASI-STATIC LOAD ON NONLINEAR... 567
(k + k − ω
1 2
2
j m1 ) φ1 j − k2 φ2 j = 0
− k φ + (k + k
2 1j 2 3 − ω 2j m2 ) φ2 j − k3 φ3 j = 0
M = M
(6)
− kl φl −1, j + ( kl + kl +1 − ω ml ) φl j − kl +1 φl +1, j
2
j = 0
M = M
− kn φn −1, j + ( kn − ω 2j mn ) φn j = 0
Solving equation (6), a general expression of the lth story stiffness can be obtain as
n mi φi j
kl = ω 2j ∑ for j = 1, 2,K , n (7)
i =1 Δφl j
Defining the damage of a structure as the reduction percentage of story stiffness before
and after damage, and assuming that the floor mass does not change due to the damage. This
damage index of each story can be expressed as
*
k / ml
DI SMR = 1− l (9)
kl / ml
For the uniform floor mass distribution and substituting the story stiffness from Eq. (7)
and Eq. (8) into Eq. (9), represented as damage index can be written as
φij*n
ω j *2 ∑
i =1 Δφlj
*
DI SMR = 1− n φ
(10)
ω j 2 ∑ ij
i =1 Δφlj
The value of this damage index can be obtained based on one set of modal parameters.
568 A. Vimuttasoongviriya, N. Kwatra and M. Kumar
F
∑φ
i =1
2
li /ωi2
DI MFDI = 1 − l* = 1 − n
(11)
∑φ
Fl *2
li /ω*2
i
i =1
where Fl and Fl * are the diagonal terms at coordinate l of the modal flexibility matrix of
undamaged and damaged structure.
3. EXPERIMENTAL PROGRAM
is shown in Figure 3.
17.5 27
Control frame Retrofitted frame
24
15
21
12.5
18
Load (kN)
Load (kN)
10 15
7.5 12
9
5
6
2.5 3
0 0
0 58
Time 0 75
Time
(a) (b)
Figure 2. Applied load history: a) control frame and b) retrofitted frame.
resonance amplitudes occur are called the natural frequencies of the structure. These
frequencies and the corresponding distribution of amplitude are global properties. Structural
mode shapes are generalized from eigenvalue problem corresponding to natural frequencies.
(a) (b)
Figure 4. Application of FRP wrap: a) flexural layers and b) confinement layers
572 A. Vimuttasoongviriya, N. Kwatra and M. Kumar
(a) (b)
Figure 6. a) An eight node brick element model and b) a schematic of FRP composite model
A shell element with 20 nodes, quadratic 3D brick element, was used to model the FRP
composite. This element allows for different material layers with different orientations. FRP
composite is that consist of two constituents. The constituents are combined at a
macroscopic level and are not soluble in each other. One constituent is the reinforcement,
which is embedded in the second constituent, epoxy, [18]. The reinforcing material is in the
form of anisotropic materials of glass fibers, which are typically stiffer and stronger than the
epoxy. Nodes of FRP layered shell elements were connected to those of adjacent concrete
solid element in order to satisfy the perfect bond assumption. Figure 6 also shows a
schematic of FRP composites. An eight-node brick element was used to model steel plates at
the supports.
between effective stress σ cef and the equivalent uniaxial strain ε eq . The equivalent uniaxial
strain is produced by the stress of concrete σ ci with modulus Eci associated with the
direction i. The equivalent uniaxial strain is calculated as
σ ci
ε eq = (12)
Eci
Unloading is a linear function to the origin. The equivalent uniaxial stress-strain diagram
for concrete and an example of the unloading point U are shown in Figure 7. The relation
between effective stress and equivalent uniaxial strain is not unique and depends on a load
history.
where Eo is the initial elastic modulus of concrete, f t 'ef is the effective tensile strength
derived from the biaxial failure function.
σ ⎧⎪ ⎛ w ⎞3 ⎫⎪ ⎛ w⎞ w
= ⎨1 + ⎜ c1 ⎟ ⎬ exp ⎜ −c2 ⎟ − (1 + c13 ) exp ( −c2 ) (14)
f t ' ef ⎪⎩ ⎝ wc ⎠ ⎪⎭ ⎝ wc ⎠ wc
where w is the crack width, wc is the crack opening at the complete release of stress,
σ is the normal stress in the crack. Values of the constants are, c1 = 3, c2 = 6.93,
respectively.
qx − x 2
σ =fef ' ef
(15)
1 + ( q − 2) x
c c
ε Eo
x= and q= (16)
εc Ec
wd
εd =εc + (17)
Ld
f c'ef = rc f c′ (18)
rc = c + (1 − c ) e
− (128 ε u )
(19)
G = rg Gc (20)
Ec
Gc = (21)
2 (1 +ν )
where rg is the shear retention factor, G is the reduced shear modulus, Gc is the initial
concrete shear modulus, Ec is the initial elastic modulus and ν is the Poisson's ratio.
A one inch thick steel plate was added at the supporting and loading location in order to
avoid stress concentration problems. At a plane of symmetry, the displacement in the
direction perpendicular to the plane was held at zero. In nonlinear analysis, the loads applied
to a finite element model are divided into a series of load increments called load steps. The
ATENA program uses Newton-Raphson equilibrium iterations for updating the model
stiffness. The monitoring points were measured at the middle of each floor, where the
largest horizontal displacements can be expected.
(a) (b)
Figure 9. FE model of retrofitted frame using FRP: a) outside view and b) inside view
12.5kN, displacement of 31 mm from initial state, DIPark of 0.49, DIk of 0.57, severe damage
state. Ultimate damage state began at the load of 16.5 kN and with a displacement of 69 mm,
leading to DIPark of 0.95 and DIk of 0.76, respectively. The ductility index was nearly 2.22
and total energy dissipation was 567.57 kN-mm.
30 30
First floor Second floor
25 25
20 20
Load (kN)
Load (kN)
15 15
10 10
(a) (b)
30
Top floor
25
20
Load (kN)
15
10
5 Retrofitted frame
Control frame
0
0 10 20 30 40 50 60 70 80 90
Displacement (mm)
(c)
Figure 10. Load-displacement plots: a) first floor, b) second floor and c) top floor.
578 A. Vimuttasoongviriya, N. Kwatra and M. Kumar
1.0 1
0.9 Retro_frame 0.9 Retro_frame
Cont_frame Cont_frame
0.8 0.8
DI (Park and Ang)
0.7 0.7
DI (Stiffness)
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
P/Pmax P/Pmax
(a) (b)
Figure 11. a) DI of Park and Ang plots and b) DI of change in stiffness plots
Figure 11 presents the correlation between applied load ratio and DIPark and DIk,
respectively. It shows that the value of DIPark of control frame has two difference trends,
while DIk seem to increase in linear. The general trend of DIPark is slightly changed because
of the cumulative effects of low deformation and energy dissipation. The high deformation
trend occurs after yield point. Based on high deformation trend, DIPark changes shape when
structure starts to lose their load carrying capacity. The value of DIk seem to increase in
linear manner because the initial stiffness ratio between undamaged and damaged structure
change slowly as compare to other damage parameters such as displacement ratio and
energy ratio. The summary of damage indexes and appearances of control frame are
presented in Table 3.
0 0 0 0 0 0 0 Un-deformed
15.0 0.91 0.65 0.61 0.25 0.27 0.26 Spalling of concrete cover
0 0 0 0 0 0 0 Un-deformed
15.0 0.60 0.49 0.19 0.12 0.17 0.18 Noise of fibre moving
21.0 0.84 0.60 0.33 0.17 0.28 0.29 Breaking noise of fibre
3m
3m
2m 2m
1m 1m
0 0
Real g
Real g
-1m
-1m
-2m
-2m
-3m
-3m
-4m
-4m 0 2 4 6 8
0 2 4 6 8 Time - s
Ti me - s
(a) (b)
Figure 12. Time history plot of control frame at top floor: a) initial state and b) final state
4 3.5 Control frame at second floor
Control frame at first floor
0 kN 0 kN
3.5 3
7.5kN 7.5kN
3 16.5kN 16.5kN
2.5
2.5
FRF (mg/N)
FRF (mg/N)
2
2
1.5
1.5
1 1
0.5 0.5
0
0
0 10 20 30 40 50
0 10 20 30 40 50
Frequency (Hz)
Frequency (Hz)
(a) (b)
6
Control frame at top floor
0 kN
5 7.5kN
16.5kN
4
FRF (mg/N)
0
0 10 20 30 40 50
Frequency (Hz)
(c)
Figure 13. FRF plots of control frame a) first floor, b) second floor and c) top floor.
EFFECT OF LATERAL QUASI-STATIC LOAD ON NONLINEAR... 581
5.2 Dynamical characteristic results and damage indexes based on modal parameters
5.2.1 Control frame
After applied each load steps, the impact hammer was used to excite the testing frame
model. The dynamic characteristics gave the records in FFT analyzer based on linear
analysis setup. These records include trigger hammer plot, time history plot and frequency
response function plot which give the amplitude of vibration along with frequency. The
analyzer was set up to make a free zoom measurement with a frequency range of 0 to 50 Hz.
In this frequency range there covered all three majority modes of this frame model. The time
history plots are shown in Figure 12 and comparison of FRF plots of control frame are
shown in Figure 13. At undamaged state the natural frequency at first, second and third
mode were 6.5Hz, 19Hz and 31.5Hz. At initial cracks of 10 kN, DIDip was 0.12, DISMR was
0.16 and DIMFDI was 0.16. At yield point of 12.5kN, the natural frequency at first, second
and third mode were 5.5Hz, 17Hz and 27.2Hz. DIDip was 0.16, DISMR and DIMFDI were 0.21,
severe damage state. At ultimate damage state, DIDip was 0.34, DISMR was 0.36 and DIMFDI
was 0.34, respectively. The frequencies at ultimate state of first, second and third mode were
5.2 Hz, 15.5Hz and 25.9 Hz. The average changed of corresponding resonance frequency
decreased approximately 18.73%.
The relation between applied loads and damage indexes based on modal parameters of
control frame are shown in Figure 16 and 17. In order to estimate the dynamic properties of
the frame under the damaged state, the measurements data after applied load are extracted to
represent the behaviour of structure. It can be seen that DIDip, DISMR and DIMFDI are small
which agree with the fact that no visible damage was reported at load less than yield point.
Meanwhile, some index values around 0.2 are obtained. In case of control frame, it is shown
that the value of damage indexes based on change in frequency and mode shape both DISMR
and DIMFDI are slightly larger than the modal plastic softening index, DIDip. In order to
estimate the modal parameter based index which compares only the natural frequency of
structure with damage or undamaged structure, DIDip is more sensitive in the case of severe
damage state or after large cracking. The results of damage indexes obtained show that the
damage of structures for both control and retrofitted frame are said to be of great concern,
when modal parameter damage indexes increase to larger than 0.20 or the structure
experienced severe damage after yield load.
582 A. Vimuttasoongviriya, N. Kwatra and M. Kumar
8m
8m
6m 6m
4m 4m
2m 2m
0 0
Real g
Real g
-2m -2m
-4m -4m
-6m -6m
-8m -8m
0 2 4 6 8 0 2 4 6 8
Ti me - s Ti me - s
(a) (b)
Figure 14. Time history plot of retrofitted frame at top floor: a) initial state and b) final state
6 7
Retrofitted frame at first floor Retrofitted frame at second floor
0kN 0kN
5 12kN 6 12kN
25kN 25kN
5
4
FRF (mg/N)
FRF (mg/N)
4
3
3
2
2
1
1
0 0
0 10 20 30 40 50 0 10 20 30 40 50
Frequency (Hz) Frequency (Hz)
(a) (b)
9 Retrofitted frame at top floor
0kN
8
12kN
7 25kN
6
FRF (mg/N)
0
0 10 20 30 40 50
Frequency (Hz)
(c)
Figure 15. FRF plots of retrofitted frame: a) first floor, b) second floor and c) top floor
EFFECT OF LATERAL QUASI-STATIC LOAD ON NONLINEAR... 583
0.4 0.4
DI (Dipasquale and Cakmak)
Retro_frame
DI (Stiffness-mass ratio)
Retro_frame
Cont_frame Cont_frame
0.3 0.3
0.2 0.2
0.1
0.1
0.0
0.0
0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pmax
P/Pmax
(a) (b)
Figure 16. a) DI of Dipasquale and Cakmak plots and b) DI of Stiffness-mass ratio plots.
0.4
Retro_frame
Cont_frame
0.3
DI (MFDI)
0.2
0.1
0.0
0.0 0.2 0.4 0.6 0.8 1.0
P/Pmax
resuls prove that, damage indexes based on modal parameters of retrofitted frame below the
curves of control frame indicating better performance. Moreover, both damage indexes
based on deformation and change in stiffness show a much acceptable accuracy correlation
with modal parameters damage indexes in general. Although damage index based on change
in modal parameters work when damage reaches severe stage but this method with the help
of impact hammer test is a useful technique to monitor health of structure without
destructing when damage is hidden within the structure behind strengthening material
layers.
(a) (b)
Figure 18. Failure mode at final state of control frame: a) experimental test and b) FE model
5.3 Failure mechanism
5.3.1 Control frame
Experimental failure modes at final state of control frame at each floor are shown in Figure
18. At yield load, first diagonal cracks started at connection joints of top floor. These
opening cracks increased when load was increasing. The second and third diagonal cracks
occurred at beam-column connections of second and first floor respectively. It indicated that
first plastic hinges were started at top floor and the other hinges were formed at second and
first floor respectively. The structure suddenly lost their load carrying capacity when cracks
crossed from tension surface though compression surface of structural cross section
elements. At final state, both diagonal and flexural cracks were occurred on structural
elements of each floor, indicating that the failure mode of control frame was combined
between shear mode and flexural mode. Only flexural failure mode was observed at column
near stub foundation. The lateral load carrying capacity of control frame was insufficient
due to non-ductile reinforcement detailing, which included no beam-column joints
transverse reinforcement. Therefore, the top and bottom bars moved in the opposite direction
under combined forces. These forces were balanced by bond stress developed between
concrete and steel bars. In such circumstances, the plastic hinges were formed by debonded
bars and connection joints lost their capacity to carry load.
EFFECT OF LATERAL QUASI-STATIC LOAD ON NONLINEAR... 585
(a) (b)
Figure 19. Failure mode of retrofitted frame: a) experimental test and b) FE model
5.4 Results of FE method and comparative analysis
Developed analytical models were validated by comparing the load-displacement results
with experimental data. For accuracy results, the stiffness damage index needs to be scale
for each damaged element. The parameters used in this index depend on the cross section of
the element and in order to use the original parameters for elements with different properties
as compared to the experimental elements. This index is modified on the basis of the
degradation of elastic modulus and applied to FE model [19]. Figure 20 shows the backbone
load-displacement plots from the FE analysis agree well with the experimental results. The
model plots were stiffer than that from experimental results. The yield load for the FE
analysis of control frame was 13 kN, which was higher than the load of 12.5kN from the
experimental results by 4%. Lastly, the ultimate load of 16 kN from model was lower than
the ultimate load of 16.5 kN from the experimental data by 3%. Similar to the control frame,
the yield load for the model of strengthened frame was 20 kN, which was higher than the
actual frame of 18 kN by 11.11%. The ultimate load for the model was 26kN, which was
586 A. Vimuttasoongviriya, N. Kwatra and M. Kumar
higher than the ultimate load of 25 kN for the actual frame by 4%. The crack patterns at final
state from FE models corresponded well with the observed failure mode of experimental
control frame and retrofitted frame after removed FRP sheets.
30
Top floor
25
20
Load (kN)
15
10
Exp_cont_frame
ATENA_cont_frame
5
Exp_retro_frame
ATENA_retro_frame
0
0 20 40 60 80
Displacement (mm)
Figure 20. Back bone load-displacement plot between experimental results and FE models.
6. CONCLUSION
Based on the performed research investigation, the following main conclusions can be
drawn:
1) Experimental results approve that the use of FRP wrapped for structural strengthening
provides increased significant lateral load capacity. The ductile behaviour of the
strengthened frame is largely restored after the FRP composite sheets are engaged.
2) The health of RC structures of both original and strengthened specimen seemed to
become unsafe when DIPark and DIk increases larger than 0.5 or dynamical damage indexes
increase larger than 0.2. In other hand, the damaged structures are unsafe when structural
stiffness reduces to more than 50% of initial state.
3) The damage indexes of strengthened frame reduce indicating better performance as
compare to the control frame. Damage indexes based on deformation and change in stiffness
show a much acceptable accuracy correlation with dynamical damage indexes in general.
4) Although the stiffness of the damaged RC structure is regained significantly by
wrapping FRP jacket but it is not able to bridge the cracks fully. The failure of the
strengthened specimen is due to breaking up of the bond between FRP layer and concrete.
5) The analytical results and crack patterns from FE method agree well with the
experimental results.
EFFECT OF LATERAL QUASI-STATIC LOAD ON NONLINEAR... 587
REFERENCES
26(2004) 497-507.
18. Kachlakev D, Miller T, Yim S. Finite element modeling of reinforced concrete
structures strengthened with FRP laminates. Final report-SPR 316, Oregon department
of transportation and Federal highway administration, 2001.
19. Lee U, Shin J. A frequency response function-based structural damage identification
method, Computers and Structures, 80(2002) 117-32.